To download a pdf file, click here.

Oxygen Shown to Be "Mystery" Electron Donor in GaN
High Pressure Experiments Resolve Long-Standing Controversy

Eugene Haller

Researchers at the Center for Advanced Materials (CAM) Electronic Materials Program have shown that the presence of oxygen atoms is responsible for the previously unexplained "n-type" character of many gallium nitride single crystals and epitaxial layers.

The use of GaN (bandgap=3.5 eV) and its alloys with InN and AlN in optoelectronic devices is growing rapidly. High efficiency blue and green LEDs are now commercially available, and blue solid state lasers have been demonstrated. In spite of this rapid progress, many of the fundamental properties of this material are still not well understood. For example, when thin films of GaN are grown by conventional metalorganic chemical vapor deposition (MOCVD) or molecular beam epitaxy (MBE), they are often much more n-type (excess electron concentration) than expected. Both n-type and p-type (excess hole concentration) layers are required for fabricating electronic devices; thus control of background n-type activity is critically important. The unintended n-type activity was first attributed to nitrogen vacancies (missing N atom in the lattice). However, recent studies have shown that these defects would not be found in sufficiently large concentrations to account for the observed effect. More recently, O, Si, and C impurities have been implicated as possible culprits.

The CAM team was able to identify oxygen as the electron donor defect in GaN through a study of the effect of an induced change in the bandgap on the electron concentration. In most donors, the electron binding energy and, therefore, the electron concentration, does not change appreciably with a change in the material's bandgap. In contrast, the electron binding energy of other, "localized," donors does change with a change in the bandgap.

The MSD team systematically increased the bandgap of a series of GaN samples by using a high pressure diamond anvil cell and monitored the electron concentration optically with Raman spectroscopy (see figure). In intentionally oxygen-doped MOCVD-grown films, the team observed a decrease in the free electron concentration at pressures above 20 GPa. This is direct proof that oxygen atoms form localized donors in GaN. In contrast, in GaN doped with Si, another n-type donor, no change in the electron concentration was found at pressures up to 27 GPa, demonstrating that Si is not a localized donor in this pressure range. The team studied a number of undoped MOCVD-grown GaN samples which exhibited unexpected n-type doping. All of these samples showed the same decrease in electron concentration at large applied pressure as the intentionally oxygen-doped sample. The team concluded that oxygen is responsible for most cases of unintentionally introduced n-type conductivity in GaN and that the sharp drop in free electron concentration above 20 GPa, characteristic for localized donors, is a unique signature for this impurity.

This MSD work shows the importance of avoiding oxygen contamination in GaN growth, particularly when attempting to make p-type material. It is interesting that the behavior of oxygen in GaN differs from its behavior in other III-V materials where it is electrically inactive. The observation that oxygen is electrically inactive at high applied pressure also means that it will be inactive in alloys of AlGaN with modest to high Al content because alloying GaN with AlN shifts the bandgap in the same way that pressure does. On the other hand, the observation that Si retains its n-type activity at high pressure establishes that it should be a good n-type dopant in AlGaN up to an Al fraction of 50% and possibly higher.